Any way to get mappings of a label encoder in Python pandas?

Question

I am converting strings to categorical values in my dataset using the following piece of code.

data['weekday'] = pd.Categorical.from_array(data.weekday).labels

For eg,

index    weekday
0        Sunday
1        Sunday
2        Wednesday
3        Monday
4        Monday
5        Thursday
6        Tuesday

After encoding the weekday, my dataset appears like this:

index    weekday
    0       3
    1       3
    2       6
    3       1
    4       1
    5       4
    6       5

Is there any way I can know that Sunday has been mapped to 3, Wednesday to 6 and so on?

score 113 · Answer 1 · answered Mar 19 '17 at 16:27

113

You can create additional dictionary with mapping:

from sklearn import preprocessing
le = preprocessing.LabelEncoder()
le.fit(data['name'])
le_name_mapping = dict(zip(le.classes_, le.transform(le.classes_)))
print(le_name_mapping)
{'Tom': 0, 'Nick': 1, 'Kate': 2}

answered Mar 19 '17 at 16:27

chinskiy

2,557
4
21
40

1

How can I deal with this for lable encoding across multiple columns? – Brian Lee Jun 11 '21 at 03:09

score 42 · Accepted Answer · answered Feb 13 '17 at 04:36

42

The best way of doing this can be to use label encoder of sklearn library.

Something like this:

from sklearn import preprocessing
le = preprocessing.LabelEncoder()
le.fit(["paris", "paris", "tokyo", "amsterdam"])
list(le.classes_)
le.transform(["tokyo", "tokyo", "paris"])
list(le.inverse_transform([2, 2, 1]))

answered Feb 13 '17 at 04:36

Algor Troy

522
5
2

1

From what I understand - le.classes_ needs to be instantiated after the desired column is transformed to get the real values, and if you use le.classes_ on say a dataframe that had more than one column encoded say with a lambda function, then le.classes_ only recovers a dictionary mapping of the last column encoded. Why is this? – Vaidøtas I. Aug 12 '19 at 06:36
1

is there any way to persist the encoder? so we could deploy it? – Dendi Handian Nov 12 '21 at 12:11
@DendiHandian You could try to save the encoder object using pickle. – Ganesh Tata Feb 22 '22 at 20:46
1

@GaneshTata Yes. I have used it on my project. – Dendi Handian Feb 25 '22 at 01:23

Abhishek · Answer 3 · 2018-04-21T06:21:15.157

7

A simple & elegant way to do the same.

cat_list = ['Sun', 'Sun', 'Wed', 'Mon', 'Mon']
encoded_data, mapping_index = pd.Series(cat_list).factorize()

and you are done, check below

print(encoded_data)
print(mapping_index)
print(mapping_index.get_loc("Mon"))

edited Apr 21 '18 at 06:21

answered May 27 '17 at 05:23

Abhishek

3,337
4
32
51

In case you don't know what distinct values are coming in data, creating list of hard coded values might leak. LabelEncoder() however deals with whatever coming in way. – Himanshu Aug 03 '17 at 19:19

score 5 · Answer 4 · answered Mar 19 '17 at 16:44

There are many ways of doing this. You can consider pd.factorize, sklearn.preprocessing.LabelEncoder etc. However, in this specific case, you have two options which will suit you best:

Going by your own method, you can add the categories:

pd.Categorical( df.weekday, [ 
    'Sunday', 'Monday', 'Tuesday', 
    'Wednesday', 'Thursday', 'Friday', 
    'Saturday']  ).labels

The other option is to map values directly using a dict

df.weekday.map({
    'Sunday': 0,
    'Monday': 1,
     # ... and so on. You get the idea ...
})

score 3 · Answer 5 · answered Feb 13 '17 at 04:28

3

First, make a categorical series:

weekdays = pd.Series(['Sun', 'Sun', 'Wed', 'Mon', 'Mon']).astype('category')

Then, inspect its "categories":

weekdays.cat.categories.get_loc('Sun')

answered Feb 13 '17 at 04:28

John Zwinck

239,568
38
324
436

Vikas Gupta · Answer 6 · 2018-05-18T22:58:45.553

3

If you have numerical and categorical both type of data in dataframe You can use : here X is my dataframe having categorical and numerical both variables

from sklearn import preprocessing
le = preprocessing.LabelEncoder()

for i in range(0,X.shape[1]):
    if X.dtypes[i]=='object':
        X[X.columns[i]] = le.fit_transform(X[X.columns[i]])

Or you can try this:

from sklearn.preprocessing import LabelEncoder

le = LabelEncoder()
data = data.apply(le.fit_transform)

Note: This technique is good if you are not interested in converting them back.

edited May 18 '18 at 22:58

answered May 14 '18 at 20:35

Vikas Gupta

1,293
1
15
21

1

but the question is precisely on the way back conversion ... :) – Mário de Sá Vera Aug 11 '21 at 15:18

score 3 · Answer 7 · answered Jul 10 '20 at 13:15

Its very simple, they have a built-in function for this.

from sklearn import preprocessing
le = preprocessing.LabelEncoder()
..
# your model steps and when you have results
..

prediction_decoded = le.inverse_transform(prediction_encoded)
print(prediction_decoded)

score 3 · Answer 8 · answered Sep 22 '21 at 00:08

I am adding my answer even after lot of answers are there to answer this OP'ed question specifically as:

If have you already label encoded your values as:

from sklearn import preprocessing
le = preprocessing.LabelEncoder()
le.fit('column-or-list-of-values')

You can get back the mapping of value to integer codes as:

dict(zip(le.classes_,range(len(le.classes_))))

score 0 · Answer 9 · answered May 02 '20 at 10:53

0

train['cat'] = train['cat'].map(list(train['cat'].value_counts().to_frame().reset_index().reset_index().set_index('index').to_dict().values())[0])

answered May 02 '20 at 10:53

Alexandr Kosolapov

153
1
4

This would be a better answer if you explained how the code you provided answers the question. – pppery May 02 '20 at 23:36

score 0 · Answer 10 · answered Jan 11 '23 at 04:49

You can create another column with the indexed values after converting from categorical to numeric. Think of it as primary key column.

#new column,index contains converted values of column weekday data['index'] = pd.Categorical.from_array(data.weekday).labels

#view the corresponding values df.groupby(['index','weekday']).first()

Any way to get mappings of a label encoder in Python pandas?

10 Answers10

Linked